Machine Translation for Multilingual Semantic Model Bootstrapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Training effective natural language processing systems to determine semantic properties like intent across multiple languages is resource-intensive, requiring significant high-quality data and human expertise, especially when supporting multiple languages.
Innovation Solution
Leveraging machine translation to transfer semantic properties from a known language to unsupported languages, either by pre-training semantic models or dynamically translating queries during processing, allowing the application of existing intent models to determine semantic properties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised machine learning is used to train semantic models for multiple languages, then the system can determine semantic properties accurately, but the time and expense required for data collection and model training increases significantly
Solution Approach 1:
The patent applies preliminary action by pre-training a semantic model on a source language corpus before needing to support additional languages. This pre-trained model serves as a foundation that can be rapidly adapted to new languages through machine translation, avoiding the need to collect and train on new data for each language from scratch.
Solution Approach 2:
The patent uses machine translation as an intermediary mechanism to bridge between the source language corpus and target language queries. Instead of directly training on target language data, the system translates target language queries to the source language, applies the pre-trained semantic model, and obtains semantic properties through this intermediate translation step.
2Adaptability or versatility
If supervised machine learning is used to train semantic models for each language separately, then the system can support multiple languages, but the expense of obtaining high-quality data for each language increases greatly
Solution Approach 1:
The patent applies universality by creating a single semantic model that can handle multiple languages through machine translation. Instead of building separate semantic models for each language, the system uses one universal model trained on the source language that can process queries from any language by translating them to the source language first.
Solution Approach 2:
The patent uses copying by translating target language queries into the source language to create equivalent input that can be processed by the existing semantic model. This allows the system to leverage the source language corpus and model for multiple languages without needing to create separate models or collect separate data for each language.
3Reliability
If human experts manually determine semantic properties for corpus expressions, then high-quality training data can be obtained, but significant time and multiple expert resources are required
Solution Approach 1:
The patent applies self-service by using the existing source language semantic model to determine semantic properties for translated queries. Instead of requiring human experts to annotate each target language corpus, the system automatically determines semantic properties by translating queries to the source language and applying the existing model, making the system self-sufficient for multi-language support.
Data Source
AI summary
Machine translation is used to leverage the semantic properties (e.g., intent) already known for one natural language for use in another natural language. In a first embodiment, the corpus of a first language is translated to each other language of interest using machine translation, and the corresponding semantic properties are transferred to the translated corpuses. Semantic models can then be generated from the translated corpuses and the transferred semantic properties. In a second embodiment, given a first language for which there is a semantic model, if a query is received in a second, different language lacking its own semantic model, machine translation is used to translate the query into the first language. Then, the semantic model for the first language is applied to the translated query, thereby obtaining the semantic properties for the query, even though no semantic model existed for the language in which the query was specified.